Search Abstracts | Symposia | Slide Sessions | Poster Sessions
When listening to real-life code-switches, neural oscillations do not differ from unilingual processing
Poster Session A - Sandbox Series, Thursday, October 24, 10:00 - 11:30 am, Great Hall 3 and 4
This poster is part of the Sandbox Series.
Leah Gosselin1, Laura Sabourin1; 1University of Ottawa
In many communities, code-switching (i.e., using more than one language during a single communication event) is common in-group behaviour. Although some bilinguals state that code-switching is the "most comfortable" way to speak, electrophysiological (EEG/ERP) studies consistently reveal that the comprehension of code-switched sentences is cognitively effortful. Some question the ecological validity of these findings, noting that previous studies almost exclusively utilize lab-made stimuli, and usually test participants in the visual modality (i.e., reading switches). The goal of the present pilot experiment was to examine the processing of authentic auditory code-switches. We examined the neural oscillations of participants listening to genuine conversations from real-life bilingual pairs. In a precursor study, 19 pairs of French-English bilinguals (couples, friends, siblings, etc.) completed a series of speech elicitation activities, giving rise to the FEBLO-Corpus (Gosselin & Sabourin, in prep). Forty passages of unstructured discussion (M=23.5 seconds, SD=4.1) were selected from FEBLOC: twenty excerpts with code-switches and twenty unilingual excerpts (ten French, ten English). The code-switched and unilingual passages did not differ in terms of length, number of utterances, amount of turn-taking, speaker age or gender, disfluencies, or rated intelligibility (all ps>.17). Nine newly recruited French-English bilinguals (M=37 years, SD=15 years; 5 women, 3 men, 1 non-binary) passively listened to the naturalistic excerpts while their continuous EEG was recorded. Resting-state brainwaves (r-EEGs) were also collected. Following pre-processing, the signal for each excerpt (item) was split into 2 second segments (50% overlap) and a Fast Fourier transform was applied to the data. The average power density for theta (4-7 Hz), alpha (8-12 Hz), lower beta (15-18 Hz), upper beta (20-30 Hz) and lower gamma (30-40 Hz) bands were extracted for each item and participant. As a preliminary analysis, mixed models were performed on each independent frequency band; they included Type (Switch vs. Unilingual), Laterality (left, middle, right) and Longitude (frontal, central, posterior) as fixed factors, as well as subject and item as random factors. Code-switched and unilingual passages did not differ in terms of elicited power density for any frequency band (no main effect of Type or interaction with topography: all ps>.34). However, a comparison of r-EEGs and passive listening EEGs (without by-item random effects) showed consistent differences across all frequency bands (theta, alpha, lower beta: ps<.0001; upper beta, lower gamma: ps<.05). In the alpha band, several self-reported participant characteristics modulated the passive listening EEGs (more so than r-EEGs). For instance, increased language entropy and more frequent switching into the majority language was related to lower alpha density (Fs>3.2, ps<.05); by contrast, switching into the minority language was linked to increased alpha density (F=3.9, p=.02). These results suggest that processing real-life code-switches is not inherently cognitively costly—rather, bilinguals' experiential traits regulate the processing of speech, whether unilingual or code-switched. However, this can only be said about the global level of auditory processing. The novel analyses methods (i.e., applying r-EEG segmenting to auditory passages) may not have been granular enough to detect condition differences at the local switching point.
Topic Areas: Multilingualism, Speech Perception